[Oliver Kiddle]

>> The output of acroread -version cannot be piped to stdout for
>> some reason, so the version number must be read from
>> ACROREAD-PREFIX/Reader/AcroVersion.  Since acroread might have
>> been started via a wrapper script, the only failsafe way of
>> determining ACROREAD-PREFIX is to look at the output of acroread
>> -help.
> 
> Unfortunately, this doesn't work on the system I tried it on, using
> Adobe's rpm. /usr/bin/acroread is a symbolic like to
> /usr/local/Adobe/Acrobat7.0/bin/acroread. Handling that involves using ls
> to get the symlink destination.
> 
> Is acroread always a script? The following works well (and very quickly)
> on any system I have access to:
>   grep '^ver=' =acroread
> 
> From a completion function, we can use $commands[$words[1]] instead
> of =acroread.

It doesn't work when acroread is a wrapper script, which it is
in my case.  (I used a wrapper instead of a symlink, because the
script needs to do some things before it starts acroread, such
as turning off core dumps (some versions of acroread 5.x used to
crash a _lot_).)  Even if I rename the script to something other
than 'acroread', I still couldn't use =acroread since the real
acroread executable isn't in my PATH. :-/

So I'll just have to resolve the symlink using readlink.  I assume
the most portable way to do this is perl, unless there is a way
to make zsh expand the link itself.  I have attached a version that
does this.  Should handle any level of symbolic links that point to
new symbolic links.

-- 
 Haakon